On 15 Jul, Simon L Peyton Jones wrote:
> I hadn't realised that your suggestion was a propos of Standard
> Haskell.
>
> I'm pretty leery about trying to agree any new libraries at this
> stage, unless someone comes up with a worked-out, and implemented,
> specification pretty quickly. The name of the Std Haskell game
> is rapid closure.
I hope it doesn't close too quickly!
> There are just tons of things that 'ought' to be in it that aren't
> going to be.
But if there are too many things missing, no one will use Standard
Haskell - it already seems as if most of the people on this list are
going to go straight to Haskell 2, which would mean that Standard
Haskell might only be used for teaching.
Maybe it's too soon to put an openURL action in (though I would have
thought this could be specified pretty easily), but I really think
that the tidying up of the relationship between the prelude and the
standard libraries is vital. For example, I can see no reason why
PreludeIO and IO should be separate. Surely it wouldn't be much work
to put it all in the IO library? As it stands, the prelude has to
refer to the library, which, I think, underlines the inconsistency:
As far as I can tell, the only prelude operations on IOError are show
and (=), so to tell what a [system raised] error is you have to import
IO. (Although you can catch errors without looking at them or tell if
a new error is the same as an earlier one!)
I'd also like to observe that fitting Haskell 2 with Standard Haskell
ought to be easier if less is in the Prelude and more in libraries.
Jon
--
Jon Fairbairn [EMAIL PROTECTED]
18 Kimberley Road [EMAIL PROTECTED]
Cambridge CB4 1HH +44 1223 570179 (pm only, please)